It is generally believed that seasonal alternation is a gradual process marked by temperature. Explored from a large data set containing 1,686,528 data points of temperature, humidity and sunshine duration, we established a seasonal dynamic model of north China. Based on the model, we discovered a turning point on the 220th day in the annual average distribution of humidity and sunshine duration, which can be used as a characteristic node to define the date of summer-autumn alternation in north China. Our results demonstrate that the alternation of summer and autumn is not a gradual process in this region, but a mutation in the annual distributions of humidity and sunshine duration, thus revealing the statistical invariance based on the local knowledge. The study also shows that humidity and sunshine duration can better reflect the climate characteristics of north China than temperature. Because the model is region-specific, the proposed method using big data can be further extended to quantitatively define other seasonal alternations and explore other climate characteristics in different regions, so as to benefit indigenous knowledge-based climate prediction.